trimseqscommand.cpp

   1 /*
   2  *  trimseqscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Pat Schloss on 6/6/09.
   6  *  Copyright 2009 Patrick D. Schloss. All rights reserved.
   7  *
   8  */
   9
  10 #include "trimseqscommand.h"
  11
  12 //***************************************************************************************************************
  13
  14 TrimSeqsCommand::TrimSeqsCommand(string option){
  15         try {
  16
  17                 abort = false;
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string AlignArray[] =  {"fasta", "flip", "oligos", "maxambig", "maxhomop", "minlength", "maxlength", "qfile", "qthreshold", "qaverage", "allfiles"};
  25
  26                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  27
  28                         OptionParser parser(option);
  29                         map<string,string> parameters = parser.getParameters();
  30
  31                         ValidParameters validParameter;
  32
  33                         //check to make sure all parameters are valid for command
  34                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  35                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  36                         }
  37
  38                         //check for required parameters
  39                         fastaFile = validParameter.validFile(parameters, "fasta", true);
  40                         if (fastaFile == "not found") { cout << "fasta is a required parameter for the screen.seqs command." << endl; abort = true; }
  41                         else if (fastaFile == "not open") { abort = true; }
  42
  43
  44                         //check for optional parameter and set defaults
  45                         // ...at some point should added some additional type checking...
  46                         string temp;
  47                         temp = validParameter.validFile(parameters, "flip", false);
  48                         if (temp == "not found"){       flip = 0;       }
  49                         else if(isTrue(temp))   {       flip = 1;       }
  50
  51                         temp = validParameter.validFile(parameters, "oligos", true);
  52                         if (temp == "not found"){       oligoFile = "";         }
  53                         else if(temp == "not open"){    abort = true;   }
  54                         else                                    {       oligoFile = temp;               }
  55
  56                         temp = validParameter.validFile(parameters, "maxambig", false);         if (temp == "not found") { temp = "-1"; }
  57                         convert(temp, maxAmbig);
  58
  59                         temp = validParameter.validFile(parameters, "maxhomop", false);         if (temp == "not found") { temp = "0"; }
  60                         convert(temp, maxHomoP);
  61
  62                         temp = validParameter.validFile(parameters, "minlength", false);        if (temp == "not found") { temp = "0"; }
  63                         convert(temp, minLength);
  64
  65                         temp = validParameter.validFile(parameters, "maxlength", false);        if (temp == "not found") { temp = "0"; }
  66                         convert(temp, maxLength);
  67
  68                         temp = validParameter.validFile(parameters, "qfile", true);
  69                         if (temp == "not found")        {       qFileName = "";         }
  70                         else if(temp == "not open")     {       abort = 0;              }
  71                         else                                            {       qFileName = temp;       }
  72
  73                         temp = validParameter.validFile(parameters, "qthreshold", false);       if (temp == "not found") { temp = "0"; }
  74                         convert(temp, qThreshold);
  75
  76                         temp = validParameter.validFile(parameters, "qaverage", false);         if (temp == "not found") { temp = "0"; }
  77                         convert(temp, qAverage);
  78
  79                         temp = validParameter.validFile(parameters, "allfiles", false);         if (temp == "not found") { temp = "F"; }
  80                         allFiles = isTrue(temp);
  81
  82                         if(allFiles && oligoFile == ""){
  83                                 cout << "You selected allfiles, but didn't enter an oligos file.  Ignoring the allfiles request." << endl;
  84                         }
  85                         if((qAverage != 0 && qThreshold != 0) && qFileName == ""){
  86                                 cout << "You didn't provide a quality file name, quality criteria will be ignored." << endl;
  87                                 qAverage=0;
  88                                 qThreshold=0;
  89                         }
  90                         if(!flip && oligoFile=="" && !maxLength && !minLength && (maxAmbig==-1) && !maxHomoP && qFileName == ""){
  91                                 cout << "You didn't set any options... quiting command." << endl;
  92                                 abort = true;
  93                         }
  94                 }
  95
  96         }
  97         catch(exception& e) {
  98                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
  99                 exit(1);
 100         }
 101         catch(...) {
 102                 cout << "An unknown error has occurred in the TrimSeqsCommand class function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 103                 exit(1);
 104         }
 105 }
 106 //**********************************************************************************************************************
 107
 108 void TrimSeqsCommand::help(){
 109         try {
 110                 cout << "The trim.seqs command reads a fastaFile and creates ....." << "\n";
 111                 cout << "The trim.seqs command parameters are fasta, flip, oligos, maxambig, maxhomop, minlength and maxlength." << "\n";
 112                 cout << "The fasta parameter is required." << "\n";
 113                 cout << "The flip parameter .... The default is 0." << "\n";
 114                 cout << "The oligos parameter .... The default is ""." << "\n";
 115                 cout << "The maxambig parameter .... The default is -1." << "\n";
 116                 cout << "The maxhomop parameter .... The default is 0." << "\n";
 117                 cout << "The minlength parameter .... The default is 0." << "\n";
 118                 cout << "The maxlength parameter .... The default is 0." << "\n";
 119                 cout << "The trim.seqs command should be in the following format: " << "\n";
 120                 cout << "trim.seqs(fasta=yourFastaFile, flip=yourFlip, oligos=yourOligos, maxambig=yourMaxambig,  " << "\n";
 121                 cout << "maxhomop=yourMaxhomop, minlength=youMinlength, maxlength=yourMaxlength)  " << "\n";
 122                 cout << "Example trim.seqs(fasta=abrecovery.fasta, flip=..., oligos=..., maxambig=..., maxhomop=..., minlength=..., maxlength=...)." << "\n";
 123                 cout << "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta)." << "\n" << "\n";
 124
 125         }
 126         catch(exception& e) {
 127                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 128                 exit(1);
 129         }
 130         catch(...) {
 131                 cout << "An unknown error has occurred in the TrimSeqsCommand class function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 132                 exit(1);
 133         }
 134 }
 135
 136
 137 //***************************************************************************************************************
 138
 139 TrimSeqsCommand::~TrimSeqsCommand(){    /*      do nothing      */      }
 140
 141 //***************************************************************************************************************
 142
 143 int TrimSeqsCommand::execute(){
 144         try{
 145
 146                 if (abort == true) { return 0; }
 147
 148                 ifstream inFASTA;
 149                 openInputFile(fastaFile, inFASTA);
 150
 151                 ofstream outFASTA;
 152                 string trimSeqFile = getRootName(fastaFile) + "trim.fasta";
 153                 openOutputFile(trimSeqFile, outFASTA);
 154
 155                 ofstream outGroups;
 156                 vector<ofstream*> fastaFileNames;
 157                 if(oligoFile != ""){
 158                         string groupFile = getRootName(fastaFile) + "groups";
 159                         openOutputFile(groupFile, outGroups);
 160                         getOligos(fastaFileNames);
 161                 }
 162
 163                 ofstream scrapFASTA;
 164                 string scrapSeqFile = getRootName(fastaFile) + "scrap.fasta";
 165                 openOutputFile(scrapSeqFile, scrapFASTA);
 166
 167                 ifstream qFile;
 168                 if(qFileName != "")     {       openInputFile(qFileName, qFile);        }
 169
 170                 bool success;
 171
 172                 while(!inFASTA.eof()){
 173                         Sequence currSeq(inFASTA);
 174                         string origSeq = currSeq.getUnaligned();
 175                         int group;
 176                         string trashCode = "";
 177
 178                         if(qFileName != ""){
 179                                 if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 180                                 else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 181                                 if(!success)                    {       trashCode += 'q';                                                               }
 182                         }
 183                         if(barcodes.size() != 0){
 184                                 success = stripBarcode(currSeq, group);
 185                                 if(!success){   trashCode += 'b';       }
 186                         }
 187                         if(numFPrimers != 0){
 188                                 success = stripForward(currSeq);
 189                                 if(!success){   trashCode += 'f';       }
 190                         }
 191                         if(numRPrimers != 0){
 192                                 success = stripReverse(currSeq);
 193                                 if(!success){   trashCode += 'r';       }
 194                         }
 195                         if(minLength > 0 || maxLength > 0){
 196                                 success = cullLength(currSeq);
 197                                 if(!success){   trashCode += 'l';       }
 198                         }
 199                         if(maxHomoP > 0){
 200                                 success = cullHomoP(currSeq);
 201                                 if(!success){   trashCode += 'h';       }
 202                         }
 203                         if(maxAmbig != -1){
 204                                 success = cullAmbigs(currSeq);
 205                                 if(!success){   trashCode += 'n';       }
 206                         }
 207
 208                         if(flip){       currSeq.reverseComplement();    }               // should go last
 209
 210                         if(trashCode.length() == 0){
 211                                 currSeq.printSequence(outFASTA);
 212                                 if(barcodes.size() != 0){
 213                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 214
 215                                         if(allFiles){
 216                                                 currSeq.printSequence(*fastaFileNames[group]);
 217                                         }
 218                                 }
 219                         }
 220                         else{
 221                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 222                                 currSeq.setUnaligned(origSeq);
 223                                 currSeq.printSequence(scrapFASTA);
 224                         }
 225                         gobble(inFASTA);
 226                 }
 227                 inFASTA.close();
 228                 outFASTA.close();
 229                 scrapFASTA.close();
 230                 outGroups.close();
 231
 232                 for(int i=0;i<fastaFileNames.size();i++){
 233                         fastaFileNames[i]->close();
 234                         delete fastaFileNames[i];
 235                 }
 236
 237                 for(int i=0;i<fastaFileNames.size();i++){
 238                         string seqName;
 239                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 240                         ofstream outGroups;
 241                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 242
 243                         while(!inFASTA.eof()){
 244                                 if(inFASTA.get() == '>'){
 245                                         inFASTA >> seqName;
 246                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 247                                 }
 248                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 249                         }
 250                         outGroups.close();
 251                         inFASTA.close();
 252                 }
 253
 254
 255                 return 0;
 256         }
 257         catch(exception& e) {
 258                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 259                 exit(1);
 260         }
 261         catch(...) {
 262                 cout << "An unknown error has occurred in the TrimSeqsCommand class function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 263                 exit(1);
 264         }
 265 }
 266
 267 //***************************************************************************************************************
 268
 269 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 270
 271         ifstream inOligos;
 272         openInputFile(oligoFile, inOligos);
 273
 274         ofstream test;
 275
 276         string type, oligo, group;
 277         int index=0;
 278
 279         while(!inOligos.eof()){
 280                 inOligos >> type;
 281
 282                 if(type[0] == '#'){
 283                         while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 284                 }
 285                 else{
 286                         inOligos >> oligo;
 287
 288                         for(int i=0;i<oligo.length();i++){
 289                                 oligo[i] = toupper(oligo[i]);
 290                                 if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 291                         }
 292
 293                         if(type == "forward"){
 294                                 forPrimer.push_back(oligo);
 295                         }
 296                         else if(type == "reverse"){
 297                                 revPrimer.push_back(oligo);
 298                         }
 299                         else if(type == "barcode"){
 300                                 inOligos >> group;
 301                                 barcodes[oligo]=index++;
 302                                 groupVector.push_back(group);
 303
 304                                 if(allFiles){
 305                                         outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 306                                 }
 307                         }
 308                 }
 309         }
 310
 311         numFPrimers = forPrimer.size();
 312         numRPrimers = revPrimer.size();
 313 }
 314
 315 //***************************************************************************************************************
 316
 317 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 318
 319         string rawSequence = seq.getUnaligned();
 320         bool success = 0;       //guilty until proven innocent
 321
 322         for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 323                 string oligo = it->first;
 324                 if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 325                         success = 0;
 326                         break;
 327                 }
 328
 329                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 330                         group = it->second;
 331                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 332                         success = 1;
 333                         break;
 334                 }
 335         }
 336         return success;
 337
 338 }
 339
 340 //***************************************************************************************************************
 341
 342 bool TrimSeqsCommand::stripForward(Sequence& seq){
 343
 344         string rawSequence = seq.getUnaligned();
 345         bool success = 0;       //guilty until proven innocent
 346
 347         for(int i=0;i<numFPrimers;i++){
 348                 string oligo = forPrimer[i];
 349
 350                 if(rawSequence.length() < oligo.length()){
 351                         success = 0;
 352                         break;
 353                 }
 354
 355                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 356                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 357                         success = 1;
 358                         break;
 359                 }
 360         }
 361
 362         return success;
 363
 364 }
 365
 366 //***************************************************************************************************************
 367
 368 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 369
 370         string rawSequence = seq.getUnaligned();
 371         bool success = 0;       //guilty until proven innocent
 372
 373         for(int i=0;i<numRPrimers;i++){
 374                 string oligo = revPrimer[i];
 375
 376                 if(rawSequence.length() < oligo.length()){
 377                         success = 0;
 378                         break;
 379                 }
 380
 381                 if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 382                         seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 383                         success = 1;
 384                         break;
 385                 }
 386         }
 387         return success;
 388
 389 }
 390
 391 //***************************************************************************************************************
 392
 393 bool TrimSeqsCommand::cullLength(Sequence& seq){
 394
 395         int length = seq.getNumBases();
 396         bool success = 0;       //guilty until proven innocent
 397
 398         if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 399         else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 400         else                                                                                            {       success = 0;    }
 401
 402         return success;
 403
 404 }
 405
 406 //***************************************************************************************************************
 407
 408 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 409
 410         int longHomoP = seq.getLongHomoPolymer();
 411         bool success = 0;       //guilty until proven innocent
 412
 413         if(longHomoP <= maxHomoP){      success = 1;    }
 414         else                                    {       success = 0;    }
 415
 416         return success;
 417
 418 }
 419
 420 //***************************************************************************************************************
 421
 422 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 423
 424         int numNs = seq.getAmbigBases();
 425         bool success = 0;       //guilty until proven innocent
 426
 427         if(numNs <= maxAmbig)   {       success = 1;    }
 428         else                                    {       success = 0;    }
 429
 430         return success;
 431
 432 }
 433
 434 //***************************************************************************************************************
 435
 436 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 437
 438         bool success = 1;
 439         int length = oligo.length();
 440
 441         for(int i=0;i<length;i++){
 442
 443                 if(oligo[i] != seq[i]){
 444                         if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 445                         else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 446                         else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 447                         else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 448                         else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 449                         else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 450                         else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 451                         else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 452                         else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 453                         else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 454                         else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 455                         else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 456
 457                         if(success == 0)        {       break;  }
 458                 }
 459                 else{
 460                         success = 1;
 461                 }
 462         }
 463
 464         return success;
 465 }
 466
 467 //***************************************************************************************************************
 468
 469 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 470
 471         string rawSequence = seq.getUnaligned();
 472         int seqLength = rawSequence.length();
 473         string name;
 474
 475         qFile >> name;
 476         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 477         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 478
 479         int score;
 480         int end = seqLength;
 481
 482         for(int i=0;i<seqLength;i++){
 483                 qFile >> score;
 484
 485                 if(score <= qThreshold){
 486                         end = i;
 487                         break;
 488                 }
 489         }
 490         for(int i=end+1;i<seqLength;i++){
 491                 qFile >> score;
 492         }
 493
 494         seq.setUnaligned(rawSequence.substr(0,end));
 495
 496         return 1;
 497 }
 498
 499 //***************************************************************************************************************
 500
 501 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 502
 503         string rawSequence = seq.getUnaligned();
 504         int seqLength = seq.getNumBases();
 505         bool success = 0;       //guilty until proven innocent
 506         string name;
 507
 508         qFile >> name;
 509         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 510         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 511
 512         float score;
 513         float average = 0;
 514
 515         for(int i=0;i<seqLength;i++){
 516                 qFile >> score;
 517                 average += score;
 518         }
 519         average /= seqLength;
 520
 521         if(average >= qAverage) {       success = 1;    }
 522         else                                    {       success = 0;    }
 523
 524         return success;
 525 }
 526
 527 //***************************************************************************************************************
 528
 529